[PATCH 0/6] Add WQ_PERCPU to alloc_workqueue users and workqueue replacement

[PATCH 0/6] Add WQ_PERCPU to alloc_workqueue users and workqueue replacement

[Marco Crivellari]

Hi,

This series collect all the workqueue changes related to Workqueue,
specifically:

- add WQ_PERCPU to alloc_workqueue() users, so to make
  explicit they are per-cpu;

- replacement of system_unbound_wq and system_wq with system_dfl_wq and
  system_percpu_wq.

Those changes didn't received feedback and the discussions can be found at
the following URLs:

- virt: acrn: Add WQ_PERCPU to alloc_workqueue users
  https://lore.kernel.org/all/20260113134406.178587-1-marco.crivellari@suse.com/

- media: synopsys: hdmirx: replace use of system_unbound_wq with system_dfl_wq
  https://lore.kernel.org/all/20260212092846.61602-1-marco.crivellari@suse.com/

- platform: cznic: turris-omnia-mcu: replace use of system_wq with system_percpu_wq
  https://lore.kernel.org/all/20251105153718.274765-1-marco.crivellari@suse.com/

- media: ddbridge: add WQ_PERCPU to alloc_workqueue users
  https://lore.kernel.org/all/20251107141052.221108-1-marco.crivellari@suse.com/

- rapidio: rio: add WQ_PERCPU to alloc_workqueue users
  https://lore.kernel.org/all/20260223103543.119278-1-marco.crivellari@suse.com/

- umh: replace use of system_unbound_wq with system_dfl_wq
  https://lore.kernel.org/all/20251107160638.328955-1-marco.crivellari@suse.com/


Thanks!

Marco Crivellari (6):
  virt: acrn: Add WQ_PERCPU to alloc_workqueue users
  media: synopsys: hdmirx: replace use of system_unbound_wq with
    system_dfl_wq
  platform: cznic: turris-omnia-mcu: replace use of system_wq with
    system_percpu_wq
  media: ddbridge: add WQ_PERCPU to alloc_workqueue users
  rapidio: rio: add WQ_PERCPU to alloc_workqueue users
  umh: replace use of system_unbound_wq with system_dfl_wq

 drivers/media/pci/ddbridge/ddbridge-core.c           | 2 +-
 drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c | 8 ++++----
 drivers/platform/cznic/turris-omnia-mcu-gpio.c       | 2 +-
 drivers/rapidio/rio.c                                | 2 +-
 drivers/virt/acrn/irqfd.c                            | 2 +-
 kernel/umh.c                                         | 2 +-
 6 files changed, 9 insertions(+), 9 deletions(-)

-- 
2.53.0

[PATCH 1/6] virt: acrn: Add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

This continues the effort to refactor workqueue APIs, which began with
the introduction of new workqueues and a new alloc_workqueue flag in:

   commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
   commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The refactoring is going to alter the default behavior of
alloc_workqueue() to be unbound by default.

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU. For more details see the Link tag below.

In order to keep alloc_workqueue() behavior identical, explicitly request
WQ_PERCPU.

Cc: Fei Li <fei1.li@intel.com>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/virt/acrn/irqfd.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/virt/acrn/irqfd.c b/drivers/virt/acrn/irqfd.c
index acf8cd5f8f8c..aab15f94166a 100644
--- a/drivers/virt/acrn/irqfd.c
+++ b/drivers/virt/acrn/irqfd.c
@@ -206,7 +206,7 @@ int acrn_irqfd_init(struct acrn_vm *vm)
 {
 	INIT_LIST_HEAD(&vm->irqfds);
 	mutex_init(&vm->irqfds_lock);
-	vm->irqfd_wq = alloc_workqueue("acrn_irqfd-%u", 0, 0, vm->vmid);
+	vm->irqfd_wq = alloc_workqueue("acrn_irqfd-%u", WQ_PERCPU, 0, vm->vmid);
 	if (!vm->irqfd_wq)
 		return -ENOMEM;
 
-- 
2.53.0

[PATCH 2/6] media: synopsys: hdmirx: replace use of system_unbound_wq with system_dfl_wq

[Marco Crivellari]

This patch continues the effort to refactor workqueue APIs, which has begun
with the changes introducing new workqueues and a new alloc_workqueue flag:

   commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
   commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The point of the refactoring is to eventually alter the default behavior of
workqueues to become unbound by default so that their workload placement is
optimized by the scheduler.

Before that to happen, workqueue users must be converted to the better named
new workqueues with no intended behaviour changes:

   system_wq -> system_percpu_wq
   system_unbound_wq -> system_dfl_wq

This way the old obsolete workqueues (system_wq, system_unbound_wq) can be
removed in the future.

Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c b/drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c
index 61ad20b18b8d..1061ab50dd64 100644
--- a/drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c
+++ b/drivers/media/platform/synopsys/hdmirx/snps_hdmirx.c
@@ -1776,7 +1776,7 @@ static void process_signal_change(struct snps_hdmirx_dev *hdmirx_dev)
 			   FIFO_UNDERFLOW_INT_EN |
 			   HDMIRX_AXI_ERROR_INT_EN, 0);
 	hdmirx_reset_dma(hdmirx_dev);
-	queue_delayed_work(system_unbound_wq,
+	queue_delayed_work(system_dfl_wq,
 			   &hdmirx_dev->delayed_work_res_change,
 			   msecs_to_jiffies(50));
 }
@@ -2238,7 +2238,7 @@ static void hdmirx_delayed_work_res_change(struct work_struct *work)
 
 		if (hdmirx_wait_signal_lock(hdmirx_dev)) {
 			hdmirx_plugout(hdmirx_dev);
-			queue_delayed_work(system_unbound_wq,
+			queue_delayed_work(system_dfl_wq,
 					   &hdmirx_dev->delayed_work_hotplug,
 					   msecs_to_jiffies(200));
 		} else {
@@ -2253,7 +2253,7 @@ static irqreturn_t hdmirx_5v_det_irq_handler(int irq, void *dev_id)
 {
 	struct snps_hdmirx_dev *hdmirx_dev = dev_id;
 
-	queue_delayed_work(system_unbound_wq,
+	queue_delayed_work(system_dfl_wq,
 			   &hdmirx_dev->delayed_work_hotplug,
 			   msecs_to_jiffies(10));
 
@@ -2518,7 +2518,7 @@ static void hdmirx_enable_irq(struct device *dev)
 	enable_irq(hdmirx_dev->dma_irq);
 	enable_irq(hdmirx_dev->det_irq);
 
-	queue_delayed_work(system_unbound_wq,
+	queue_delayed_work(system_dfl_wq,
 			   &hdmirx_dev->delayed_work_hotplug,
 			   msecs_to_jiffies(110));
 }
-- 
2.53.0

[PATCH 3/6] platform: cznic: turris-omnia-mcu: replace use of system_wq with system_percpu_wq

[Marco Crivellari]

Currently if a user enqueues a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.

This lack of consistency cannot be addressed without refactoring the API.

This patch continues the effort to refactor worqueue APIs, which has begun
with the change introducing new workqueues and a new alloc_workqueue flag:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

Replace system_wq with system_percpu_wq, keeping the same behavior.
The old wq (system_wq) will be kept for a few release cycles.

Cc: Marek Behun <kabel@kernel.org>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/platform/cznic/turris-omnia-mcu-gpio.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/platform/cznic/turris-omnia-mcu-gpio.c b/drivers/platform/cznic/turris-omnia-mcu-gpio.c
index 7f0ada4fa606..4e430d6c3fc4 100644
--- a/drivers/platform/cznic/turris-omnia-mcu-gpio.c
+++ b/drivers/platform/cznic/turris-omnia-mcu-gpio.c
@@ -893,7 +893,7 @@ static bool omnia_irq_read_pending_old(struct omnia_mcu *mcu,
 
 	if (status & OMNIA_STS_BUTTON_PRESSED) {
 		mcu->button_pressed_emul = true;
-		mod_delayed_work(system_wq, &mcu->button_release_emul_work,
+		mod_delayed_work(system_percpu_wq, &mcu->button_release_emul_work,
 				 msecs_to_jiffies(FRONT_BUTTON_RELEASE_DELAY_MS));
 	} else if (mcu->button_pressed_emul) {
 		status |= OMNIA_STS_BUTTON_PRESSED;
-- 
2.53.0

[PATCH 4/6] media: ddbridge: add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

This continues the effort to refactor workqueue APIs, which began with
the introduction of new workqueues and a new alloc_workqueue flag in:

   commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
   commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The refactoring is going to alter the default behavior of
alloc_workqueue() to be unbound by default.

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU. For more details see the Link tag below.

In order to keep alloc_workqueue() behavior identical, explicitly request
WQ_PERCPU.

Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/media/pci/ddbridge/ddbridge-core.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/media/pci/ddbridge/ddbridge-core.c b/drivers/media/pci/ddbridge/ddbridge-core.c
index 40e6c873c36d..d240e291ba4f 100644
--- a/drivers/media/pci/ddbridge/ddbridge-core.c
+++ b/drivers/media/pci/ddbridge/ddbridge-core.c
@@ -3430,7 +3430,7 @@ int ddb_init_ddbridge(void)
 
 	if (ddb_class_create() < 0)
 		return -1;
-	ddb_wq = alloc_workqueue("ddbridge", 0, 0);
+	ddb_wq = alloc_workqueue("ddbridge", WQ_PERCPU, 0);
 	if (!ddb_wq)
 		return ddb_exit_ddbridge(1, -1);
 
-- 
2.53.0

[PATCH 5/6] rapidio: rio: add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

This continues the effort to refactor workqueue APIs, which began with
the introduction of new workqueues and a new alloc_workqueue flag in:

   commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
   commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The refactoring is going to alter the default behavior of
alloc_workqueue() to be unbound by default.

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU. For more details see the Link tag below.

In order to keep alloc_workqueue() behavior identical, explicitly request
WQ_PERCPU.

Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Alexandre Bounine <alex.bou9@gmail.com>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/rapidio/rio.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/rapidio/rio.c b/drivers/rapidio/rio.c
index 0c175e6d424f..f45e58b0971f 100644
--- a/drivers/rapidio/rio.c
+++ b/drivers/rapidio/rio.c
@@ -1994,7 +1994,7 @@ int rio_init_mports(void)
 	 * TODO: Implement restart of discovery process for all or
 	 * individual discovering mports.
 	 */
-	rio_wq = alloc_workqueue("riodisc", 0, 0);
+	rio_wq = alloc_workqueue("riodisc", WQ_PERCPU, 0);
 	if (!rio_wq) {
 		pr_err("RIO: unable allocate rio_wq\n");
 		goto no_disc;
-- 
2.53.0

[PATCH 6/6] umh: replace use of system_unbound_wq with system_dfl_wq

[Marco Crivellari]

This patch continues the effort to refactor workqueue APIs, which has begun
with the changes introducing new workqueues and a new alloc_workqueue flag:

   commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
   commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The point of the refactoring is to eventually alter the default behavior of
workqueues to become unbound by default so that their workload placement is
optimized by the scheduler.

Before that to happen, workqueue users must be converted to the better named
new workqueues with no intended behaviour changes:

   system_wq -> system_percpu_wq
   system_unbound_wq -> system_dfl_wq

This way the old obsolete workqueues (system_wq, system_unbound_wq) can be
removed in the future.

Cc: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/all/20250221112003.1dSuoGyc@linutronix.de/
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 kernel/umh.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/umh.c b/kernel/umh.c
index cffda97d961c..48117c569e1a 100644
--- a/kernel/umh.c
+++ b/kernel/umh.c
@@ -430,7 +430,7 @@ int call_usermodehelper_exec(struct subprocess_info *sub_info, int wait)
 	sub_info->complete = (wait == UMH_NO_WAIT) ? NULL : &done;
 	sub_info->wait = wait;
 
-	queue_work(system_unbound_wq, &sub_info->work);
+	queue_work(system_dfl_wq, &sub_info->work);
 	if (wait == UMH_NO_WAIT)	/* task has freed sub_info */
 		goto unlock;
 
-- 
2.53.0

Re: [PATCH 3/6] platform: cznic: turris-omnia-mcu: replace use of system_wq with system_percpu_wq

[Marek Behún]

Acked-by: Marek Behún <kabel@kernel.org>

On Tue, May 05, 2026 at 02:40:59PM +0200, Marco Crivellari wrote:
> Currently if a user enqueues a work item using schedule_delayed_work() the
> used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
> WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
> schedule_work() that is using system_wq and queue_work(), that makes use
> again of WORK_CPU_UNBOUND.
> 
> This lack of consistency cannot be addressed without refactoring the API.
> 
> This patch continues the effort to refactor worqueue APIs, which has begun
> with the change introducing new workqueues and a new alloc_workqueue flag:
> 
> commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
> commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")
> 
> Replace system_wq with system_percpu_wq, keeping the same behavior.
> The old wq (system_wq) will be kept for a few release cycles.
> 
> Cc: Marek Behun <kabel@kernel.org>
> Suggested-by: Tejun Heo <tj@kernel.org>
> Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
> ---
>  drivers/platform/cznic/turris-omnia-mcu-gpio.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/cznic/turris-omnia-mcu-gpio.c b/drivers/platform/cznic/turris-omnia-mcu-gpio.c
> index 7f0ada4fa606..4e430d6c3fc4 100644
> --- a/drivers/platform/cznic/turris-omnia-mcu-gpio.c
> +++ b/drivers/platform/cznic/turris-omnia-mcu-gpio.c
> @@ -893,7 +893,7 @@ static bool omnia_irq_read_pending_old(struct omnia_mcu *mcu,
>  
>  	if (status & OMNIA_STS_BUTTON_PRESSED) {
>  		mcu->button_pressed_emul = true;
> -		mod_delayed_work(system_wq, &mcu->button_release_emul_work,
> +		mod_delayed_work(system_percpu_wq, &mcu->button_release_emul_work,
>  				 msecs_to_jiffies(FRONT_BUTTON_RELEASE_DELAY_MS));
>  	} else if (mcu->button_pressed_emul) {
>  		status |= OMNIA_STS_BUTTON_PRESSED;
> -- 
> 2.53.0
> 

Re: [PATCH 0/6] Add WQ_PERCPU to alloc_workqueue users and workqueue replacement

[Tejun Heo]

Hello,

These are equivalent conversions that are part of tree-wide conversion. I'll
route these through the workqueue tree in a day or two. Please holler if
there are any objections.

Thanks.

-- 
tejun

Re: [PATCH 0/6] Add WQ_PERCPU to alloc_workqueue users and workqueue replacement

[Frederic Weisbecker]

Le Tue, May 05, 2026 at 07:16:25AM -1000, Tejun Heo a écrit :
> Hello,
> 
> These are equivalent conversions that are part of tree-wide conversion. I'll
> route these through the workqueue tree in a day or two. Please holler if
> there are any objections.

Thanks a lot!

> 
> Thanks.
> 
> -- 
> tejun

-- 
Frederic Weisbecker
SUSE Labs

Re: [PATCH 0/6] Add WQ_PERCPU to alloc_workqueue users and workqueue replacement

[Marco Crivellari]

On Tue, May 5, 2026 at 7:16 PM Tejun Heo <tj@kernel.org> wrote:
>
> Hello,
>
> These are equivalent conversions that are part of tree-wide conversion. I'll
> route these through the workqueue tree in a day or two. Please holler if
> there are any objections.
>
> Thanks.

Many thanks!

-- 

Marco Crivellari

SUSE Labs